Imaging features reliably distinguish malignancy in periosteal cartilaginous tumors: a two-center analysis of 89 patients.

OBJECTIVES: Periosteal cartilaginous tumors, comprising periosteal chondroma (PCh) and periosteal chondrosarcoma (PChS), are uncommon bone surface tumors that remain poorly characterized radiologically. This study aimed to identify clinical, and imaging features distinguishing PChS from PCh. METHODS: A retrospective analysis was conducted on patients with histologically-confirmed PCh and PChS from two sarcoma reference centers (2001-2024). Inclusion required at least one imaging modality (X-ray, CT or MRI). Radiological features depicting the tumor dimensions, matrix, and peritumoral changes (including periosteal reaction, peritumoral edema and enhancement patterns) were collected. Tumor shape was quantified using shape radiomics. Uni- and multivariable analyses (including area under the ROC [AUROC] analysis, Student, Wilcoxon, Chi-square tests, and stepwise binary logistic regression) were used to identify predictors of PChS. RESULTS: 89 patients were included (median age: 31 years, 32.6% PChS). Several variables were associated with PChS: longest diameter (LD) ≥5cm (P < 0.0001), volume ≥19cm3 (P < 0.0001), flat bone and humerus locations (P < 0.0001), aggressive periosteal reaction (P = 0.0002), Lodwick classification (P = 0.0004), peritumoral soft-tissue edema (P = 0.0128) and contrast-enhancement (P = 0.0306). Among shape radiomics features, tumor volume achieved the highest AUROC (0.923) but did not outperform LD (AUROC=0.903, DeLong P = 0.65). Stepwise multivariable analysis identified four variables associated with malignancy: tumor size, location, peritumoral soft-tissue edema, and aggressive periosteal reaction. Among these, LD ≥ 5 cm (OR = 76.29, P < 0.0001) and location (humerus: OR = 21.68, P = 0.0239; flat bones: OR = 32.63, P = 0.0475) were independent predictors. Diagnostic performance was highest when at least two of these four features were present (accuracy=93.1%), with perfect specificity when three were present. CONCLUSIONS: Imaging can reliably distinguish benign from malignant periosteal cartilaginous tumors, whereas shape radiomics provides no meaningful incremental value over simple tumor size assessment. ADVANCES IN KNOWLEDGE: Integrating clinical, CT, and MRI features provides high diagnostic accuracy in distinguishing benign from malignant periosteal cartilaginous tumors.